Hi, On Wed, 26 Apr 2023 15:37:58 +0100, Conor Dooley wrote: > Perhaps the thing to do is to actually take Yangyu's first patch and my > second one, since the problem with backwards compatibility doesn't stop > the kernel from being more permissive? How about taking my first patch[1] since the ECR[2] mentioned that the format of the ISA string is defined in the RISC-V unprivileged specification? However, I think we still need to stop the parser if some characters that the parser is not able to handle as Andrew Jones mentioned in the previous mail[3]. In this case, we still need to add some code to stop parsing if any error happens. In my humble opinion, backward compatibility can be solved by backports to LTS kernels. I think the better option is to allow using uppercase letters in the device-tree document to make the parser coherent with RISC-V ISA specification but recommend using all lowercase letters for better backward compatibility. [1] https://lore.kernel.org/all/tencent_63090269FF399AE30AC774848C344EF2F10A@xxxxxx/ [2] https://drive.google.com/file/d/1nP3nFiH4jkPMp6COOxP6123DCZKR-tia/view [3] https://lore.kernel.org/all/d7t6nxmblmyqriogs4bxakpse3qc7pc6cczjnhmkpk4kjwvgcb@3aihh3erdn6p/ Thanks, Yangyu Chen
